Nonfeasance and the End of Policy?

Reflections on the Revolution in Public

Authority Liability
Jonathan Morgan
Senior Lecturer in Law, Fellow of Corpus Christi College, University of Cambridge

Introduction

For at least forty years English tort lawyers have debated the proper approach to the
duty of care in negligence. Since Caparo Industries plc v Dickman,1 it has been
orthodox to approach a novel claim by proceeding cautiously from established
situations of liability and weighing up the factors for and against liability. These
factors include the relationship between the parties (‘proximity’) and all other relevant
factors (ie policy, under the rubric ‘fair, just and reasonable’). In Robinson v Chief
Constable of West Yorkshire Police the Supreme Court has again explained this
approach—and its limits.2 Policy reasoning has been declared unnecessary and even
impermissible within ‘established categories’—and the scope of what should be taken
as ‘established’ was widely conceived. So nonfeasance was categorised as
established ‘non-liability’, whereas direct infliction of physical harm was, conversely,
an ‘established’ category where liability invariably arises. In neither situation,
therefore, was the Caparo approach to be taken, according to the majority judgment
of Lord Reed. No ‘novel’ issue was said to arise.
For public authorities, the great majority of claims involve allegations that
they carelessly failed to protect the claimant from harm (although Robinson itself
concerned positive acts). Thus a very wide-ranging absence of liability has been
confirmed. But it is now said to depend entirely on the ‘omissions principle’ and not
on any consideration of the possible negative consequences of negligence liability.
Robinson therefore raises a number of elementary propositions about liability
for negligence. It has far-reaching implications for public authority liability—and
beyond. On one view the Supreme Court’s decision that a duty of care was owed on
the facts seems so obvious, and the Court of Appeal’s denial of duty such an
aberration, that there is little more to say. But as the difference between the two
appellate courts (each unanimous) suggests, deciding whether to classify a case as a
‘hard’ or ‘easy’ one (i.e. whether liability is ‘established’ or not) can itself be hard.
Even elementary propositions have ragged edges—and reasonable (indeed extremely
senior) judges may disagree about when axioms run out.
As will be seen, the three Supreme Court judgments in Robinson take different
approaches to these central issues. Not everyone will be convinced that the majority’s
dichotomy between acts and omissions captures all the relevant considerations in this
vexed area of negligence. It is a radical departure from three decades of case-law and
requires us to treat the reasoning in many decisions at the highest level as having been
misconceived—indeed beside the point. So what? A decision that ‘fits’ badly with
earlier precedents could be a great advance if the innovation explains the law better.

1
[1990] 2 AC 605.
2
[2018] UKSC 4; [2018] 2 WLR 595 (‘Robinson’).
But can the basic idea of ‘no liability for omissions’ do so? The readiness with which
exceptions to this ‘omissions principle’ can be judicially created raises doubts. These
are examined later in the article.
Another concern is that if overt consideration of factors for and against
liability (‘policy’) is prohibited it may nevertheless continue to exert unacknowledged
influence over the development of the formal (‘principled’) legal categories. This
would reduce the transparency of legal development. Judges must fully explain their
reasons, which are thereby exposed to public scrutiny. Let us accept that there is no
need for elaborate policy reasoning in routine negligence cases. Also, policy
reasoning can be done badly and perhaps it often is. The solution is to criticise bad
policy reasoning and call for a more rigorous approach. Seeking to prohibit it
altogether across a very wide class of cases risks throwing out baby, bathwater, cradle
and all.
Baroness Hale and Lord Hodge concurred with Lord Reed in Robinson. Lord
Mance and Lord Hughes delivered their own separate judgments. While they agreed
that there should be liability on the facts, Lord Mance and Lord Hughes repudiated
Lord Reed’s desiccated approach to the duty of care in general. Lord Mance accepted
(as any lawyer must) that there is no need to ‘resort’ to the Caparo approach in every
established negligence case.3 But he proceeded to argue nonetheless that the
recognition, application and development of ‘established’ liabilities depends on
reasons of policy. Lord Mance specifically stated that police negligence cases could
not satisfactorily be understood using the acts-omissions distinction alone; they were
difficult to fit within ‘any absolutely fixed legal mould’. 4 Similarly, Lord Hughes
began by agreeing that it is ‘neither necessary nor appropriate’ to go through the full
Caparo approach in every negligence case, ie when liability clearly arises in a
‘recurrent factual situation’.5 But Lord Hughes was emphatic that the ‘omissions
principle’ was not ‘the only, or sufficient, reason’ for the absence of police liability.
Policy reasons were necessary to bolster that conclusion.6
How is it that senior judges disagree on the basic task for the court faced with
a dispute about the existence of a duty of care? We begin with the basic principles.
In ‘tort law on the back of a postage stamp’ the ‘test’ for duty of care was laid down
by the House of Lords in Caparo v Dickman. Of course this is too simplistic. It is,
first, a serious misunderstanding to suggest that Caparo has to be cited in every
negligence case. In practice there is no doubt that a duty is owed in the vast bulk of
routine negligence litigation concerning drivers, employers and doctors. It is beyond
argument that such defendants owe duties of care to other road-users, employees and
patients. There are numerous other well-established duty situations. In all these cases
the question of duty (or its absence) is determined simply by precedent. It is only
when a novel claim presents itself that the court must return to consider liability from
first principles. All the judges agreed on this in Robinson—but not on the crucial
question of what precisely is a ‘novel’ case.
While Caparo is certainly the orthodox approach in such novel cases, it is
another serious mischaracterisation to view it as a ‘test’ in any true sense. There are
two weighty reasons. First, given the wide variety of situations that may arise and the
infinite permutation of factors for and against liability that they involve, it is a priori
unlikely that a brief formula—even one split into three elliptical parts like
3
Robinson [83].
4
Robinson [88] (and generally [84]-[94]).
5
Robinson [100].
6
Robinson [114] (and generally [113]-[119]).
‘foreseeable; proximate; fair, just and reasonable’—could ever hope to resolve the
question. It is quite impossible. Secondly, it must always be emphasised that the law
lords in Caparo clearly recognised that impossibility. It is a puzzling irony that Lord
Bridge’s speech is taken to have inaugurated the tripartite ‘test’ even though he said
this in the very paragraph from which ‘the test’ is taken:

the concepts of proximity and fairness embodied in these additional

ingredients are not susceptible of any such precise definition as would be
necessary to give them utility as practical tests, but amount in effect to little
more than convenient labels to attach to the features of different specific
situations which, on a detailed examination of all the circumstances, the law
recognises pragmatically as giving rise to a duty of care of a given scope.7

It is welcome that in Robinson the Supreme Court again stressed that although
Caparo provides a broad framework for the court’s inquiry, it is in no meaningful
sense a ‘test’.8 The worrying thing is that this elementary point has had to be made so
often.9
We have now advanced to ‘tort on the back of a postcard’: distinguishing
between established and novel cases in negligence; with Caparo the general approach
to novel claims but certainly not ‘a test’. For added points the student of tort law
would note that the reasoning in Caparo emphasised the desirability of incremental
steps from decided cases. This seems not just incontestable but downright banal. Is
this not how the common law works—in every conceivable context? 10 That the point
had to be laboured in Caparo was, of course, a reaction to the expansionary effect of
Anns v Merton London Borough Council.11 Under Lord Wilberforce’s approach a
duty of care arose whenever harm to the claimant was reasonably foreseeable (which
means virtually every case in practice), absenting countervailing policy reasons. The
Anns question ‘why not?’ was superseded in Caparo by decidedly more cautious
rhetoric, ‘why should there be liability?’, using decided cases as the starting point.
Caparo was, and is, about reigning in the expansionary tendencies of negligence.
Shortly before Caparo Lord Goff made the following observation:

the broad general principle of liability for foreseeable damage is so widely

applicable that the function of the duty of care is not so much to identify cases
where liability is imposed as to identify those where it is not.12

7
[1990] 2 AC 605, 617-618.
8
Robinson [24] per Lord Reed (quoting the same passage) (‘ironic’).
9
See further J Morgan, ‘The rise and fall of the general duty of care’ (2006) 22 Prof Neg 206,
discussing Customs and Excise Commissioners v Barclays Bank plc [2006] UKHL 28, [2007]
1 AC 181.
10
eg Southern Pacific Co v Jensen, 244 US 205, 221 (1917) per Holmes J: ‘judges do and
must legislate, but they can do it only interstitially; they are confined from molar to molecular
actions’. The cricketer might recall Hirst’s recommendation to ‘get them [the winning runs]
in singles’—rather than attempting to hit the ball for six.
11
[1978] AC 728, 751-752.
12
Smith v Littlewoods Organisation [1987] AC 241, 280-281.
In the Court of Appeal in Robinson Hallett LJ suggested a complementary view of
Caparo itself: ‘an attempt by the House of Lords to impose a degree of control over
the growth of the law of negligence generally’.13
A final elementary proposition concerns constitutional law. As AV Dicey
famously wrote, agents of the state (for example the police) are liable for torts that
they commit in the same way as anybody else—unless they can show specific lawful
authority for what with otherwise be tortious.14 This is the crucial legal
presupposition underlying battery and false imprisonment claims against the police. 15
It is no less true in negligence. Can it be doubted that police drivers owe duties to
other road users to take reasonable care even when (eg) pursuing a fleeing criminal?
The Police Federation is currently campaigning for an exemption from civil liability
for ‘emergency response drivers’.16 That legislation would be necessary to secure
such exemption underlines the Diceyan point. It is, of course, trite law that it may not
necessarily be careless to take risks in an emergency situation.17 Again though, the
premise is that a duty to take reasonable care exists; appropriate (perhaps less
stringent) care must still be taken in emergencies.18
How did these elementary principles of negligence law apply, with such
controversy, in Robinson?

Robinson: Facts and Decision

Elizabeth Robinson, aged 76, was walking down a busy street in Huddersfield when
she was barged into by two policemen attempting to arrest a drug dealer (who was
struggling to resist). All four fell to the ground and Mrs Robinson was injured. In her
claim for damages the county court recorder held that the police had been negligent
(viz the sergeant in command should have awaited the arrival of reinforcement before
attempting the arrest). However, the recorder held that the police owed no duty of
care and dismissed the claim. Mrs Robinson’s first appeal failed. The Court of
Appeal agreed that there was no duty of care owed and also reversed the recorder’s
finding of breach of duty.19 However the Supreme Court reversed both grounds of the
Court of Appeal’s decision in turn, and remitted Mrs Robinson’s claim to the trial
judge for assessment of damages.
The Court of Appeal had decided that even though Robinson involved direct
infliction of foreseeable physical injury it was necessary to consider the Caparo
factors. Doing so, the Court of Appeal held that the reasons counting against police
13
[2014] EWCA Civ 15 [9].
14
eg Introduction to the Study of the Law of the Constitution, 3rd ed (1889), p 181 (quoted in
Robinson [32]).
15
And others: ‘Trespass trips up the zealous bureaucrat, the eager policeman and the officious
citizen’: T Weir, A Casebook on Tort (10th ed Sweet & Maxwell 2004) 322.
16
Emergency Response Drivers (Protections) Bill (First Reading, House of Commons, 19
December 2017 (10 Minute Rule)), Clause 2. See also Clause 3 (reduced sentences for
Dangerous Driving) and Clause 4 (immunity from prosecution for Careless Driving). NB
speed limits are already inapplicable to fire, police and ambulance vehicles: Road Traffic
Regulation Act 1984, s 87.
17
eg Daborn v Bath Tramways [1946] 2 All ER 333; Watt v Hertfordshire CC [1954] 1 WLR
835; King v Sussex Ambulance Service NHS Trust [2002] EWCA Civ 953, [2002] ICR 1413
(cf Buxton LJ at [46]-[50]); Social Action, Responsibility and Heroism Act 2015.
18
eg Marshall v Osmond [1983] QB 1034.
19
[2014] EWCA Civ 15, [2014] PIQR P14.
liability in the House of Lords decisions in Hill, Brooks and Van Colle likewise
excluded a duty of care in Robinson.20 But Lord Reed firmly disagreed. The Caparo
approach is appropriate only in novel cases. Barging into a fellow pedestrian is
however, he held, a straightforward and obvious case for a duty of care. No elaborate
consideration of policy reasons against liability was necessary, nor indeed
permissible, in such an established duty situation.
Moreover, he held, the supposed entrenchment of the ‘defensive policing’
concern (relied upon by the Court of Appeal) was an illusion. Lord Reed agreed that
the leading cases cited above had been correctly decided. That was because they had
each involved omissions (failures by the police to protect the claimants). As Michael v
Chief Constable of South Wales had shown,21 that the police were guilty only of
nonfeasance was quite sufficient to show that no duty of care arose. There was no
need to rely on ‘defensive policing’ to deny liability in these omissions cases—and
there never had been.
That Lord Keith had famously invoked ‘defensive policing’ in Hill was for
Lord Reed explicable by the Anns approach then prevailing in negligence (ie that
liability prima facie arose for foreseeable failures to act—as in Anns itself—absenting
policy reasons why not). Given the contemporary Anns doctrine, it had been
necessary for the court in Hill to adduce policy reasons to rebut the prima facie
liability for omissions. But since Anns had been overruled,22 and the ‘no omissions
liability’ rule reaffirmed,23 the ‘defensive policing’ argument could no longer be
viewed as essential to the decisions correctly holding police forces not liable for
omissions. It was certainly not to be applied to the situation in Robinson where, in
accordance with Diceyan equality before the law, police officers owed the same
duties of care to fellow pedestrians as anybody else. Finally, the Court of Appeal had
been wrong to overturn the recorder’s decision that the police had carelessly breached
that duty.
At least at first sight the Supreme Court’s decision seems undeniably correct.
It takes an admirably straightforward approach to the duty of care. It stirringly
vindicates an important aspect of equality before the law. However, as Lord Mance’s
and Lord Hughes’s judgments suggest, the straightforwardness might be more
apparent than real. Recall Lord Rodger’s observation about formulating tests for the
duty of care: ‘the unhappy experience with the rule so elegantly formulated by Lord
Wilberforce in [Anns] suggests that appellate judges should follow the philosopher’s
advice to “Seek simplicity, and distrust it”.’24
In particular three concerns arise. First whether a policy-free zone of ‘easy
cases’ involving ‘established liability’ can be marked out as readily as Lord Reed
suggests. Secondly, whether the revisionist approach downplaying ‘defensive
policing’ is sustainable as a matter of precedent and is anyway desirable. Thirdly,
whether the elevation of the acts-omissions distinction into dispositive watershed
status may not occasion much litigation, since the clarity of that distinction dissolves

20
Hill v Chief Constable of West Yorkshire [1989] AC 53; Brooks v Commissioner of Police
of the Metropolis [2005] UKHL 24; [2005] 1 WLR 1495; Van Colle v Chief Constable of
Hertfordshire; Smith v Chief Constable of Sussex [2008] UKHL 50, [2009] 1 AC 225.
21
[2015] UKSC 2; [2015] A.C. 1732.
22
Murphy v Brentwood DC [1991] 1 AC 398.
23
Stovin v Wise (Norfolk CC, Third Party) [1996] AC 923; Gorringe v Calderdale MBC
[2004] UKHL 15; [2004] 1 WLR 1057.
24
Customs v Barclays Bank [2006] UKHL 28, [51].
on examination. The risk is of unstable over-simplicity (or as it was once put in a
different legal context, ‘uncertain certainty’).25

‘Established categories’: No need for policy reasoning?

As stated above, there is of course no need to engage in elaborate Caparo-style

reasoning when a duty of care is well established. This is exemplified by stock
negligence litigation (such as road, workplace and medical accidents). The same is
true, presumably, for situations where the absence of a duty of care is well
established. In such cases the duty of care is a non-issue.
But what about claims which could plausibly be characterised as requiring an
extension or retrenchment of ‘established liability’? How wide, precisely, should the
‘established categories’ themselves be drawn? Often this will be controversial. Such
controversy is inherent in the common law method. Is a given dispute governed
squarely by precedent or should earlier authorities be distinguished, or extended by
analogy? It is an elementary point of common law technique that the extent of
binding precedent requires identification of the rationale for the rule previously down
and whether that rationale properly extends to the dispute before the court. It may
well be in the forensic interest of one of the parties to attempt to distinguish (or
extend) past precedent. Where the weight of precedent is heavy and longstanding, it
will be hard to convince the court to depart from it. But if a sufficiently strong
argument can be adduced, the court may yet do so. Precedent ensures the common
law’s stability; but there is always potential for development too. Such development
usually occurs through incremental development, refinement and gloss. Precedent is
not an iron cage. Deciding what the common law rule actually is may often require
exercise of judgement. Even when a rule is utterly clear, the court should be willing
to hear the argument of a party with a powerful enough reason to suggest a departure
from precedent.
In negligence specifically, there will very rarely be proper occasion to revisit
the long-established duty of care in (say) an everyday road accident claim. A student
suggesting a first-principles Caparo approach would make a howler; a lawyer trying
to argue that a driver does not owe a duty of care would presumably receive short
shrift from the court (perhaps even a wasted costs order). We can trust that courts
would not lightly permit futile rehearsal of unarguable points. Can it not be left at
that? Must there further be bright-line rules about when courts are allowed to hear
novel arguments (for the extension, or retrenchment, of liability)? For this even to
feasible there would have to be absolute categories of liability. But these do not exist.
For example, it has often been suggested that while liability is recognised cautiously
for certain exotic kinds of harm (eg pure economic loss, psychiatric injury) and for
harm caused by indirect means (eg by statements, by omission, by shock), in the core
case of directly inflicted physical injury a duty of care is automatic. Liability here is
trite. This rule seems almost absolute.
But the final qualification (‘almost’) is important. The courts have not
previously deemed it impermissible to raise cogent arguments against liability for
directly inflicted physical damage. In the Marc Rich case the House of Lords held
that Caparo applies even in such cases.26 (On the facts, the defendant classification
society was held not to owe a duty of care when it certified as seaworthy a ship which

25
FA Mann (1981) 97 LQR 379.
26
Marc Rich & Co AG v Bishop Rock Marine (The Nicholas H) [1996] AC 211.
sank with the loss of the claimant’s cargo.) For the majority, Lord Steyn expressly
rejected the submission that when there is physical damage to the claimant’s property
the only criterion for duty is that such damage was reasonably foreseeable. While
accepting that ‘the qualitative difference between cases of direct physical damage and
indirect economic loss’ was material, Lord Steyn nevertheless held the Caparo
approach ‘relevant to all cases whatever the nature of the harm sustained by the
plaintiff’.27 Relying upon a combination of policy factors, he decided accordingly that
no duty was owed by the classification society.
In solitary dissent in Marc Rich, Lord Lloyd found these policy factors
unconvincing. More fundamentally he denied their relevance. Lord Lloyd believed:
‘All that is required is a straightforward application of Donoghue v. Stevenson’ (ie,
reasonable foreseeability).28 The further qualifications of ‘proximity’ and ‘fair, just
and reasonable’ had been developed (in Caparo) ‘as a means of containing liability
for pure economic loss’. However: ‘None of these difficulties arise in the present
case. We are not here asked to extend the law of negligence into a new field. We are
not even asked to make an incremental advance’.29 For Lord Lloyd, Marc Rich was a
straightforward case of physical property damage. A duty of care arose automatically.
But of course, his dissent contrasts sharply with the majority decision.
In Robinson the Court of Appeal relied heavily on the Marc Rich case. Lord
Steyn’s reasoning unequivocally showed that ‘policy’ may potentially be relevant in
‘direct physical damage’ cases. In Hallett LJ’s view, the ‘fair, just and reasonable’ (ie
policy) component of Caparo—

may have been triggered by the desire to constrain the development of the law
of negligence in relation to claims which do not involve direct physical
damage, but it has become part of the general law. In the vast majority of
claims the answer to the question posed at the third stage of the test—whether
it is fair just and reasonable to impose a duty—may be obvious but it still
applies.30

In other words, while liability ‘very often goes without saying’ in physical damage
cases (to quote Lord Lloyd in Marc Rich),31 that is not the same thing as saying that
such liability can never be rebutted by a sufficiently strong policy reason.
It seems impossible to reconcile with Marc Rich an absolute rule that
foreseeable physical damage is always actionable, rendering policy reasons
inadmissible as a potential means of rebutting liability. Marc Rich is dealt with only
very briefly by Lord Reed in Robinson, who did not suggest the decision was
incorrect. He sought to sidestep the awkward precedent. Lord Reed first said that the
reasoning in Marc Rich ‘was essentially directed to considerations relevant to
economic loss’.32 With respect, that is untenable. Lord Reed’s suggestion does not do
justice even to the reasoning of Lord Lloyd who was (in dissent) the closest of the law
lords to Lord Reed’s ‘established category’ approach—but for that very reason Lord
Lloyd stressed that Marc Rich was not a case of pure economic loss. The claimant in
27
Ibid 235.
28
Ibid 230.
29
Ibid 229-230
30
[2014] EWCA Civ 15 [40].
31
[1996] AC 211, 230. Lord Lloyd went on to accept that ‘exceptional’ circumstances (only)
might preclude a duty of care.
32
Robinson [28]
Marc Rich was the owner of the cargo destroyed through the defendant’s alleged
negligence.
Secondly, Lord Reed endorsed the decision in Perrett v Collins which
subdivided the ‘physical damage’ category between damage to property and personal
injury.33 In Perrett the Court of Appeal declined to follow Marc Rich, finding a duty
of care in an otherwise similar situation involving bodily injury to the claimant.
Robinson similarly involved personal injuries. But full analysis of the merits of Marc
Rich and Perrett v Collins would have been welcome. The distinction between
property damage and personal injury drawn in the later case, which many would
approve, awaits authoritative consideration.
Even assuming the distinction in Perrett v Collins to be sound, it still does not
support an absolute rule that direct personal injury is always actionable. If that were
true the decision in Mulcahy v Ministry of Defence must be wrong.34 The claimant
soldier suffered personal injury when a howitzer gun was fired, allegedly negligently,
before he moved a safe distance from it. The incident occurred during the Gulf War
in 1991. The Court of Appeal held that soldiers do not owe duties of care to each
other to avoid personal injury during armed combat. This decision has never, to our
knowledge, been questioned.35 It was accepted as correct by the Supreme Court in
Smith v MoD.36 But if it is now forbidden even to entertain reasons for denying a duty
of care in direct personal injury cases, was Mulcahy wrongly decided?
Lord Reed’s final observation on Marc Rich was that it had been ‘mistaken’ to
believe that ‘the Caparo case established a tripartite test’ for the duty of care. As seen
above, Caparo did no such thing, nor could it. But the House of Lords in Marc Rich
did not make the schoolboy error of supposing Caparo to be a mechanical test. The
court fully assessed the policy reasons advanced for and against liability—Caparo
was applied properly in Marc Rich.
Lord Reed’s failure to engage fully with the reasoning of the Marc Rich case
is unfortunate given the reliance on it by the court below. Indeed in his judgment in
Robinson, Lord Mance too cited a long passage from Marc Rich and pointed out that
Lord Steyn had been ‘cautious to warn against absolute rules in that respect’ (ie
concerning direct infliction of physical damage). 37 Lord Mance emphasised the
following words of Lord Steyn’s opinion: ‘in most cases of the direct infliction of
physical loss or injury through carelessness, it is self-evident that a civilised system of
law should [recognise] a duty of care’.38 Most, but not all.
Lord Mance made the point explicit. While accepting that Caparo is
irrelevant in well-established negligence situations, the key question becomes whether
a given claim concerns ‘established’ liability or not. Lord Mance found this as a
considerably more difficult question than Lord Reed. For example, even for direct
physical damage resulting from positive conduct Lord Mance was ‘not persuaded that
it is always a safe guide at the margins’ (having accepted that in general there would
be liability); Lord Mance expressly doubted whether (as Lord Reed thought) direct

33
Robinson [28] citing Perrett v Collins [1999] PNLR 77, 92 per Hobhouse LJ.
34
[1996] QB 732. I am grateful to Dr Janet O’Sullivan for emphasising the importance of
this case in the present context.
35
For PS Atiyah, Mulcahy was ‘surely entitled to the prize for the most undeserving damages
claim of the decade (which is saying something)’: The Damages Lottery (1997) 90.
36
[2013] UKSC 41, [2014] AC 52.
37
Robinson [90]
38
[1996] AC 211, 235.
physical damage constitute ‘axiomatically such a category [of established liability],
whatever the precise circumstances’.39
Lord Mance did not agree that Robinson itself required merely a
straightforward application of established liability. Rather:

there was open to the law a genuine policy choice whether or not to hold the
police responsible on a generalised basis for direct physical intervention on the
ground, causing an innocent passer-by physical injury, in the performance of
their duties to investigate, prevent and arrest for suspected offending by some
third person(s). In my opinion, that policy choice should now be made
unequivocally in the sense indicated by Lord Reed.40

Lord Mance therefore agreed with the Court of Appeal that policy was relevant; he
did not think that the precedents for liability were determinative as they were not
‘particularly close to the present case’ (and inconsistent with one later authority). 41
But in the end, differing from the Court of Appeal, Lord Mance preferred Lord
Keith’s statement in Hill that ‘a police officer, like anyone else, may be liable in tort
to a person who is injured as a direct result of his acts or omissions’; 42 and this
extended to ‘the direct physical interface between the police and the public, in the
course of an arrest placing an innocent passer-by or bystander at risk’. 43 So Lord
Mance concurred in the result in Robinson, but this should not disguise the divergent
reasoning. There is a considerable difference between considering the arguments for
and against liability and making the ‘policy choice’ in favour (for Lord Mance on
Diceyan rule of law grounds) and declaring that because liability is ‘established’ there
is no argument to be had.
The different emphasis is carried across into the treatment of policy more
generally. Lord Reed accepted that policy reasons (‘exercise of judgment about the
potential consequences of a decision’) do still ‘sometimes have a role to play’ when
the court is invited to recognise a novel duty. 44 He also warned that courts ‘are not
policy-making bodies in the sense in which that can be said of the Law Commission
or government departments’.45 As seen, Lord Reed deemed policy reasoning
impermissible in established (or ‘non-novel’) cases. Yet for Lord Mance, even
accepting that ‘The courts are not a Law Commission’,46 the importance of policy
reasoning should not be sidelined. Policy choices by generations of courts have
created the ‘established’ liabilities found in the law today—it must be so unless we
sign up to the ‘fairy tale’ in which ‘the common law has not changed since the Saxon
era, merely to be revealed from time to time by an increasingly perceptive judiciary’.
Instead, as Lord Mance said, ‘in recognising the existence of any generalised duty in
particular circumstances [the courts] are making policy choices’.

39
Robinson [85].
40
Robinson [95].
41
Robinson [96] citing Knightley v Johns [1982] 1 WLR 349, Rigby v Chief Constable of
Northamptonshire [1985] 1 WLR 1242 and Desmond v Chief Constable of Nottinghamshire
[2011] EWCA Civ 3; [2011] PTSR 1369.
42
[1989] AC 53, 59
43
Robinson [97].
44
Robinson [42], [69](2).
45
Robinson [69](2)
46
Robinson [84]
 Whatever the jurisprudential truth of how the common law is made, there are
practical constraints. As Hobhouse LJ said in another remark cited by Lord Reed, for
claims falling within ‘established categories of liability, a defendant should not be
allowed to seek to escape from liability by appealing to some vaguer concept of
justice or fairness; the law cannot be re-made for every case’. 47 That is not to deny
that policy reasoning has shaped the law. But as Lord Bingham said in Rees v
Darlington, refusing to entertain reconsideration of the McFarlane case decided ‘as
recently as four years ago’, it would do no credit to the law if cases were overruled
because a different bench of judges would weigh the competing policy considerations
differently.48
But taken to extremes the proper desire for the law to remain settled and
certain would be a recipe for ossification. Even Lord Reed accepted that the Supreme
Court may depart from established authority; however he apparently thought that
lower courts should not reconsider the policy determinations inherent in established
categories, not even the Court of Appeal. A somewhat more liberal approach was
suggested in Jones v Kaney by Lord Dyson towards denials of duty of care. In Lord
Dyson’s view denying remedies to the victims of wrongs requires ‘strict and cogent
justification’ which ‘should be kept under review’ by the courts. 49 Unjustified
immunities may undermine public confidence in the law. This suggests that claimants
should always be permitted to argue that established areas of ‘non-liability’ should be
reconsidered. But if defendants produce convincing evidence that negligence liability
is causing harmful consequences, should the courts not equally be entitled to
reconsider an established category of liability (or making exceptions to it)? Should
existing liabilities not also be ‘kept under review’? Can development of the common
law really be for the Supreme Court alone?50
If the Robinson approach persists we can expect much argument in the lower
courts about whether cases are ‘novel’—as the touchstone for the propriety of hearing
policy arguments about the duty of care. A court incorrectly classifying an
‘established’ duty (or immunity) as raising a ‘novel’ question faces being told by the
Supreme Court not just that it incorrectly weighed up the relevant policy concerns, but
erred in considering them at all. Following Robinson the Supreme Court has again
rebuked the Court of Appeal for incorrectly treating as novel a claim falling ‘squarely
within an established category of duty of care’. 51 It is surprising that senior judges in
the Court of Appeal have twice been criticised for making category mistakes in their
reasoning—the sort of error that inexperienced students might make when labouring
through Caparo to establish a duty of care in a standard motoring case. The majority

47
Perrett v Collins [1999] PNLR 77, 90–91 (cited Robinson [26]).
48
Rees v Darlington Memorial Hospital NHS Trust [2003] UKHL 52, [2004] 1 AC 309 [7].
Cf McFarlane v Tayside Health Board [2000] 2 AC 59.
49
Jones v Kaney [2011] UKSC 13, [2011] 2 AC 398 [113]-[114] citing Rondel v Worsley
[1969] 1 AC 191, Saif Ali v Sydney Mitchell & Co [1980] AC 198 and Hall (Arthur JS) & Co
v Simons [2002] 1 AC 615.
50
NB further Lord Reed’s view that the Supreme Court’s criteria to determine ‘quasi-
employment’ in vicarious liability should not be assessed for ‘fairness, justice and
reasonableness’ when applied to new cases: Cox v Ministry of Justice [2016] UKSC 10,
[2016] AC 660 [41]. Similarly per Lord Reed in Armes v Nottinghamshire CC [2017] UKSC
60, [2018] AC 355 [36] (non-delegable duty) (an implied rebuke to policy reasoning below by
Males J, Burnett LJ (now Lord Chief Justice) and Black LJ (now Lady Black JSC)).
51
Darnley v Croydon NHS Trust [2018] UKSC 50, [2018] 3 WLR 1153, [15]-[16] (waiting
time information given by casualty department receptionist).
in Robinson (and now the unanimous Supreme Court in Darnley) had a more
expansive view of ‘established’ liability than others hitherto would. Did the Court of
Appeal really commit such elementary blunders of common law technique? The
Supreme Court’s intention seems to be to discourage policy reasoning, envisaging
disposal of cases by adherence to precedent where possible. As will now be
discussed, whether that is desirable is open to question.

Irrelevance of ‘Defensive Policing’?

A prominent and controversial policy argument deployed against negligence liability

is the prediction of ‘defensive behaviour’. It is frequently claimed that far from
encouraging the defendant to be more careful, liability will result in damaging
defensive practices. A common anecdotal example is the claim that US doctors
perform millions of clinically unnecessary diagnostic tests annually, to preclude
negligence suits for failing to detect latent conditions. In England however the
argument is more commonly raised in public authority cases. In its classical form the
premise is that public authorities are already doing their best to serve the public in
difficult circumstances (ie constrained budgets which require difficult choices
between competing priorities). The incentives of negligence liability would not be
merely benign (‘be more careful!’) but would distract authorities from their task.
Negligence would distort the priorities that would otherwise be settled in the general
public interest (eg divert resources towards groups more likely to claim). 52 As Lord
Keith put it, ‘the cure may be worse than the disease’.53
This sounds plausible. But that such damaging consequences will happen (or
do happen) is very far from proven by evidence. Given its plausible-but-unproven
status the ‘defensive administration’ argument has unsurprising enjoyed a chequered
career in English law. Frequently it has been rejected at the highest level. 54 But in
one line of authority at least, negligence claims against the police, it was consistently
accepted by the courts—until Michael and now Robinson. Courts have long been
aware that hard evidence is needed to resolve the debate. Lord Phillips CJ feared in
Van Colle that evaluating the impact of liability on the police was ‘not easy’ and ‘not
readily resolved by a court of law’.55 The Lord Chief Justice’s relief that the Law
Commission had ‘just published a Consultation Paper … that directly addresses the
issues’ was perhaps overstated since the Law Commission did little more than collect
the extant empirical research in an appendix; it did not undertake any further research
of its own. Ultimately the Commission declined to make substantive proposals for
reform of public authority liability.56 So the issue remains unsettled.
52
NB the prediction that negligence liability to those reporting threats of violence would
reorient police priorities ‘to treat these particular reported threats with especial caution at the
expense of the many other threats to life, limb and property’: Van Colle [132] per Lord
Brown. See further eg Mitchell v Glasgow CC [2009] UKHL 11, [2009] 1 AC 874 [28] per
Lord Hope.
53
Rowling v Takaro Properties Ltd [1988] AC 473, 502.
54
eg Home Office v Dorset Yacht Co [1970] AC 1004, 1033 per Lord Reid (‘Her Majesty’s
servants are made of sterner stuff’); Spring v Guardian Assurance Plc [1995] 2 AC 296
(reference-writers); Capital & Counties plc v Hampshire CC [1997] QB 1004, 1043-4 per
Stuart-Smith LJ (fire brigade); Darnley [22] (casualty department receptionists).
55
Van Colle [102]
56
Ibid; Law Commission Consultation Paper No 187, Administrative Redress: Public Bodies
and the Citizen (2008), Appendix B; cf Law Commission Report No 322 (2010).
 Various commentators have said that the courts should stop relying on
‘defensive administration’ as a reason to deny liability. It is an empirical claim (or
prediction) without the evidence necessary to substantiate it. Sir Basil Markesinis and
Jörg Fedtke argue that when ‘defensive administration’ has been so influential for
public authority liability it is both surprising and worrying that the evidence for it is
‘so thin’ (‘a dearth of empirical data’); tort law should therefore ‘leave the quicksand
of speculation’.57 Specifically the authors criticise Lord Hoffmann’s prediction that
liability for highway authorities would distort public spending priorities towards road
improvements, leaving less for education or social services (‘it is important … to
consider the cost to the community of the defensive measures which [local
authorities] are likely to take in order to avoid liability’). 58 Markesinis and Fedtke
state that in the absence of empirical evidence Lord Hoffmann’s claim is ‘just
speculation’—and ‘Legal arguments cannot be solved, and litigation cannot be
determined, on the basis of hunches, however eminent and experienced their source
may be’.59 Perhaps stung by the criticism, in a subsequent lecture Lord Hoffmann
reported that he had undertaken private research into the effects of the decision that
local authority psychologists owe duties of care to children undergoing educational
assessment.60 Having interviewed the Chief Educational Psychologist of
Leicestershire County Council Lord Hoffmann reached a clear conclusion: ‘I saw
enough to suggest [Phelps] has been an unqualified disaster for this country’s
education system. It has produced about three trivial and probably aberrant
[compensation] awards at enormous cost [for defendant authorities], both financially
and emotionally’.61 Lord Hoffmann’s main point was to call for experience-based
cost-benefit analysis of all proposed new liabilities for public authorities.
The debate seems intractable. The possibility of real and significant harm
flowing from liability cannot be ruled out. But nor can it be demonstrated on the
available evidence (everybody agrees that authoritative studies should be carried out;
but nobody actually does them). How should the law respond? Two divergent paths
exist. First the ‘defensive administration’ argument could be rejected as unfounded
speculation. Or secondly it could be considered despite the lack of definitive proof,
by analogy perhaps with the ‘precautionary principle’ central to environmental law
and policy.62 If the (alleged) distortion of public-authority priorities and/or the
(alleged) defensive behaviour of front-line staff can be viewed as sufficiently
damaging to the public interest, it would be rational to take action against a plausible
threat of such socially undesirable consequences, even in the absence of ‘full
scientific certainty’ (ie incontrovertible empirical proof that it will definitely happen).
Not surprisingly, support for each approach can be drawn from the English
jurisprudence. In Perrett v Collins Buxton LJ refused counsel’s invitation to
speculate about the effect of liability upon aircraft classification societies (in the
admitted absence of evidence). Buxton LJ said that courts should be ‘very cautious’
57
BS Markesinis and J Fedtke, ‘Authority or Reason? The Economic Consequences of
Liability for Breach of Statutory Duty in a Comparative Perspective’ [2007] EBLR 5, 51-52.
58
Stovin v Wise [1996] AC 923, 958.
59
[2007] EBLR 5, 26. Cf Michael [160] per Lord Kerr (‘instinctual reactions’).
60
Phelps v Hillingdon LBC [2001] 2 AC 619.
61
Lord Hoffmann, ‘Reforming the Law of Public Authority Negligence’ (Bar Council Law
Reform Lecture 2009).
62
eg Rio Declaration on Environment and Development (1992), Principle 15: ‘Where there
are threats of serious or irreversible damage, lack of full scientific certainty shall not be used
as a reason for postponing cost-effective measures to prevent environmental degradation’.
before deciding such issues without the assistance of expert witnesses.63 But not all
courts are so circumspect. In Caparo v Dickman the House of Lords denied a duty of
care in part because indeterminate liability would ‘open up a limitless vista of
uninsurable risk’ for the defendant auditors.64 The House of Lords seems to have
accepted the defendants’ submissions to this effect. 65 Yet in the Court of Appeal
(where the majority had recognised a duty of care to existing shareholders), Bingham
LJ said that the insurance argument was ‘hard to assess in the absence of evidence or
inquiry’ and Taylor LJ pointedly observed that ‘no evidence was adduced’ to support
the defendants’ arguments that insurance would be ‘difficult and cripplingly
expensive’ or that accountants might even decline to act as auditors.66
Yet courts often make predictions about the effect of liability. For purists this
is unacceptable. Some commentators reject policy reasoning in tort law altogether. 67
While that would require a radical revision of longstanding practice, nobody
welcomes loose or speculative reasoning. It could be that when the court suspects
that liability might have detrimental effects, but without knowing for certain, it should
proceed to recognise liability as ‘principle’ requires and leave it to Parliament to
overturn this if the consequences are unacceptable. That was Taylor LJ’s position
about the supposedly uninsurable liability in Caparo.68 The most famous example of
this strategy is Lord Scarman’s speech in McLoughlin v O’Brian.69 However, Lord
Edmund-Davies emphatically reaffirmed the relevance of policy in the same case
(provided it was established with ‘clarity and cogency’ and given ‘close scrutiny’).70
Lord Bridge, although alongside Lord Scarman in doubting the cogency of the
policies limiting recovery for nervous shock in McLoughlin v O’Brian, rejected a
possible approach of leaving it to the legislature to extend a policy-limited sphere of
liability; for that would be ‘an unwarranted abdication of the court’s function of
developing and adapting principles of the common law to changing conditions’.71
These are choppy jurisprudential waters with no land in sight. What precisely
are the limits of the judicial function? Is it realistic to leave policy to the legislature
exclusively? Would this result in two distinct laws of tort—statutory and common
law? Would statute, driven by assessment of public policy, eventually engulf a
common law characterised by simple principles followed through to their conclusions,
heedless of the practical consequences?72 The decision in Robinson provides a novel
contribution to the ceaseless debate by avoiding it. That at least was Lord Reed’s
conclusion for the majority, although it failed to convince Lord Hughes and Lord
Mance.
63
[1999] PNLR 77, 114-115.
64
[1990] 2 AC 605, 643 per Lord Oliver.
65
Ibid 609-611.
66
[1989] QB 653, 689, 703.
67
eg R Stevens, Torts and Rights (Oxford University Press, 2007) 306-314; A Beever,
Rediscovering the Law of Negligence (Hart Publishing, 2007) 171-173, 176.
68
[1989] QB 653, 703.
69
[1983] 1 AC 410, 430
70
[1983] 1 AC 410, 426-428.
71
[1983] 1 AC 410, 441.
72
Lord Scarman said elsewhere that the common law, although ‘delightful’ was ‘now of
marginal importance only’ and ‘can play only a peripheral part in keeping the law in touch
with modern developments’ (Hansard, HL Debates, 15 December 1982, col 634). Lord
Scarman’s approach in McLoughlin v O’Brian would surely accelerate statutory replacement
of the common law.
 Lord Reed’s position has seductive clarity. If the police directly injure
somebody (when driving, dashing around on foot, or otherwise), they owe a duty of
care as would anyone else. But when the police have failed to protect the claimant,
they do not owe a duty of care, any more than any other person would. This sharp
act-omission dichotomy will be explored below. Its relevance here is that it permitted
Lord Reed to sideline the ‘defensive policing’ reasoning in the leading cases of Hill,
Brooks and Van Colle. As noted already, Lord Reed commented that Lord Keith’s
influential remarks in Hill were unnecessary for the decision reached in that case, as
now understood. The police were rightly under no duty of care because they had no
positive duty to protect Miss Hill from harm. The same was true in Brooks and Van
Colle, as in Michael (where Lord Toulson deliberately declined to rely on ‘defensive
policing’).73 The decisive point in all these cases was the doomed allegation that the
police owed a positive duty to protect the claimants. That and not public policy was
the key to the jurisprudence on police liability. As Lord Reed put it, the ‘ultimate
reason’ against liability ‘does not depend merely on a policy devised by a recent
generation of judges in relation to policing: it is based on the application of a general
and long-established principle’ of the common law—no liability for nonfeasance.74
But the House of Lords had emphatically approved the ‘defensive policing’
argument (dubbed the ‘core principle’ from Hill’s case) in Brooks and Van Colle.
Lord Reed explained away Hill as a product of the Anns era, when rebutting prima
facie liability for nonfeasance required countervailing policy reasons. That may be
true for Hill itself. But Brooks (2005) and Van Colle (2008) postdate not only the
overruling of Anns in Murphy v Brentwood DC (as Lord Hughes observes),75 but also
postdate Stovin v Wise and Gorringe v Calderdale. Stovin and Gorringe were,
according to Lord Reed, the key decisions by which the confusion from Anns (in this
area) was ‘finally dissipated’ and the law on omissions liability returned to
orthodoxy.76 Is it really plausible that the House of Lords in Brooks and Van Colle
overlooked those leading cases and so continued ‘cite authorities from [the Anns v
Merton] period without always appreciating the extent to which their reasoning has
been superseded’?77
Given how closely together these cases were decided there was inevitably
some overlap of judicial personnel. Lord Nicholls (who sat in Brooks) had been a
member of the House of Lords in Stovin v Wise, and both Lord Steyn and Lord
Rodger (who also sat in Brooks) had been in Gorringe v Calderdale. Lord Brown sat
in Van Colle and Brooks and Gorringe. These law lords can hardly have overlooked
the lessons of the ‘nonfeasance cases’ when relying on the ‘defensive policing’
argument in Brooks and Van Colle. The policy-free ‘orthodoxy’ which Lord Reed
claims the nonfeasance cases represent is also open to question. As seen above, Lord
Hoffmann supported his leading decision in Stovin v Wise with concerns about
defensive administration. Can anyone believe that Lord Hoffmann made these
observations entirely gratuitously, having misunderstood the implications of the
nonfeasance rule that he was influentially restating?
In Robinson Lord Hughes was unpersuaded by Lord Reed’s revisionist stance.
In his view, the various statements of policy in numerous cases (especially Hill,
Brooks and Van Colle) were ‘simply too considered, too powerful and too
73
Michael [121]-[122].
74
Robinson [69](1); eg East Suffolk Rivers Catchment Board v Kent [1941] AC 74.
75
Robinson [113]
76
Robinson [31].
77
Ibid.
authoritative in law to be consigned to history’. 78 Lord Hughes thought the ‘defensive
administration’ concern had been central to those decisions and not merely
‘supporting arguments’. The reluctance of the common law to impose general
liability for omissions was correct but insufficient by itself. Moreover, Lord Hughes
found the policy reasons cogent; he advocated their continued use not only on
grounds of precedent:

it is sometimes asserted that that part of the policy considerations which

related to the danger of defensive policing lacks hard evidence. That may
technically be so, since there has not existed the kind of duty of care which
would test it in practice. But like Lord Brown in [Van Colle] I for my part
would regard that risk as inevitable. It can scarcely be doubted that we see the
consequences of defensive behaviour daily in the actions of a great many
public authorities. I do not see that it can seriously be doubted that the threat
of litigation frequently influences the behaviour of both public and private
bodies and individuals.79

Lord Mance showed similar scepticism in his Robinson judgment. On Hill

and Van Colle he said that the cases ‘can, I agree, be rationalised as cases of omission,
but that was not how they were reasoned’.80 Lord Mance cited various other cases in
which courts had relied on concerns similar to ‘defensive administration’ without
drawing any distinction between acts and omissions.81 Thus Lord Mance also views
Lord Reed’s approach as conceptual rewriting of the leading cases. Denial of liability
has frequently rested on ‘defensive administration’ and not on the distinction between
acts and omissions.
Even the European Court of Human Rights seems to have accepted the
‘defensive policing’ argument. In Osman v United Kingdom the European Court held
that a high degree of police incompetence in responding to threats of violence would
be needed to establish a breach of Article 2, ECHR (right to life). This was not made
out on the facts.82 The European Court justified a high threshold for breach by the
imperative to avoid ‘an impossible or disproportionate burden on the authorities’
faced with ‘difficulties involved in policing modern societies, the unpredictability of
human conduct, [and] the operational choices which must be made in terms of
priorities and resources’.83 As Lord Phillips CJ says, ‘Those factors could be said to
reflect, to a degree, some of the policy considerations identified in Hill’.84 (It appears
that Article 2 liability has expanded significantly since Osman and that its cautionary
words have perhaps been overlooked by the European Court during that expansion.85)
Whether the 3-2 majority in Robinson can reverse the English courts’ previous
‘considered, powerful, authoritative’ reliance on ‘defensive administration’ remains to

78
Robinson [113].
79
Robinson [112]
80
Robinson [88]
81
eg [86] (Smith v Ministry of Defence [2014] AC 52), [89] (Elguzouli-Daf v Metropolitan
Police Commissioner [1995] 1 QB 335).
82
(1998) 29 EHRR 245
83
Ibid [116]
84
Van Colle [91]
85
DSD v Metropolitan Police Commissioner [2018] UKSC 11; [2018] 2 WLR 895, [127] per
Lord Hughes.
be seen. Much will depend on whether the acts-omissions distinction can bear the
weight that Lord Reed’s approach places upon it. There is room for doubt.

Acts and Omissions: The Great Divide?

As set out above, in Robinson Lord Reed held that many of the leading cases denying
public authority liability should now be recognised as straightforward applications of
the nonfeasance principle. Lord Reed argued there is no need to use policy reasoning
to explain the authorities (and to the extent that courts had previously done so, policy
should no longer be treated as necessary for their decisions). Thus a simple
dichotomy will govern future public authority negligence claims. Failures to act will
generate liability only exceptionally. By contrast, positive misfeasance leading to
physical harm will ordinarily sound in liability (subject to causation, breach and
defences such as statutory authority).86 This promises to simplify a long-vexed area.
But the dichotomy is fuzzier than it sounds.
That this involves a radical re-interpretation of the authorities has been noted
already. For example in Brooks Lord Steyn had expressly rejected counsel’s attempt
to confine the Hill approach to nonfeasance cases. Lord Steyn said the proposed
distinction was ‘unmeritorious’ and ‘hardly does justice to the essential reasoning in
Hill's case’ which had been applied in cases of ‘alleged positive and direct negligence
by the police’.87
It is dangerous to forecast the future. But there is reason to think that policy
reasons will still bear upon the acts/omissions distinction. For the distinction is more
controversial, and less clear, than our summary of Lord Reed’s approach suggests.
This is so for a number of reasons. First, many have questioned whether ‘Diceyan
logic’ must entail that public authorities possess the same freedom to do nothing
(faced with dangers and threats) that private individuals enjoy. The very premise of
the nonfeasance doctrine (for public authorities) is debateable. Secondly, the
omissions principle is not absolute. Of course Lord Reed acknowledges that positive
duties to act can arise where (eg) the defendant created the danger from which the
claimant needs protection, or where the defendant voluntarily assumed responsibility
for the claimant’s safety. Both of those exceptions is of rather uncertain width. The
very coherence of ‘voluntary assumption of responsibility’ is questionable. The
exceptions cannot determine the outcome of cases in uncontroversial fashion,
precluding the need to take other relevant matters into account. We consider these
criticisms in turn.
First the ‘omissions principle’ itself, and its application to public authorities.
Lord Hoffmann explained the rule that there is no liability for nonfeasance by
considerations of personal autonomy.88 It would be considerably more intrusive to
require those doing nothing to take positive steps to improve the position of others
(compared to the misfeasance rule, that one must take care not to harm others when
carrying on some activity). Many would accept this argument for private individuals
—although it is certainly not beyond challenge when, as frequently noted, it diverges
from our moral duties to aid those in distress (the very point of the biblical parable

86
See Gorringe [3] per Lord Steyn: ‘in a case framed in negligence, against the background
of a statutory duty or power, a basic question is whether the statute excludes a private law
remedy’ (emphasis in original).
87
Brooks [32].
88
Stovin v Wise [1996] AC 923, 943-944.
from which Lord Atkin’s neighbour principle was derived). 89 But Lord Hoffmann’s
argument seems less convincing for public bodies (although expounded in a claim
against a highway authority). Should governmental bodies really have ‘personal
freedom’ (to do nothing) entrenched by tort law? Is their literal raison d’être not to
use the powers and resources conferred upon them to safeguard the public? The
Stovin reasoning has been influentially questioned in the police context by Stelios
Tofaris and Sandy Steel,90 although their critique was accepted only by the dissenting
judges in the leading case of Michael v Chief Constable where the nonfeasance rule
was reaffirmed. 91
Undoubtedly that rule is entrenched as a matter of positive law. The ‘moral’
reasoning in Stovin provides questionable foundations for it. But such ‘principled’
arguments did not stand alone. As seen in the previous section, Lord Hoffmann also
offered policy reasons for caution. Other policy reasons could be suggested to
reinforce the omissions rule. For example, liability might distract an authority from
making decisions in the public interest (eg if the general good required certain
individuals’ interests to be neglected or even sacrificed).92 The undesirability of
‘conflict of duties’ has been accepted in other areas of negligence.93 As seen, the
argument that police priorities might be ‘distorted’ by negligence liabilities was
prominent in Van Colle. The House of Lords accepted squarely that individual
claimants’ interest in recovering damages had to be subordinated to the public good. 94
So pace Lord Reed the acts-omissions distinction could well depend in part on ‘policy
devised by a recent generation of judges’, not just ‘general and long-established
principle’.
Another consequential argument against liability for omissions proceeds from
the extensive liability that public authorities would otherwise bear. It is scarcely an
exaggeration to say that few accidents ever occur which could not have been
prevented by some agency of the state intervening to address the risk in question—
whether by policing, inspecting, regulating or taking some other protective action.
New laws could be passed and new agencies set up to provide new powers of
intervention when existing powers, extensive as they are, appear insufficient. It is a
familiar theme of the contemporary public (media) discourse surrounding any tragedy
that the government is to blame for not having prevented it—and that steps must
rapidly be taken to ensure it never happens again. (Once more this suggests the
political unreality of Lord Hoffmann’s notion of ‘freedom to do nothing’ extended to
public authorities.) The reason for tort law to be cautious in the face of such public
expectations is captured by Cherie Booth QC and Dan Squires:

without a distinction between acts and omissions, the state can be held
responsible for any failure to improve society. How such improvement can
best be achieved, or even what is regarded as an improvement at all, are
questions we ordinarily think of as being best left to the political process.
89
Noted eg Mitchell [39] per Lord Scott (Pharisees not liable in English law). For criticism
eg J Kortmann, Altruism in Private Law (Oxford University Press, 2005).
90
‘Negligence Liability for Omissions and the Police’ [2016] CLJ 128.
91
Michael [177] per Lord Kerr, [197] per Baroness Hale (both dissenting).
92
eg Capital & Counties at 1036 (Duke of York pulling down houses to contain the Great
Fire of London).
93
eg JD v East Berkshire NHS Trust [2005] UKHL 23; [2005] 2 AC 373, James-Bowen v
Metropolitan Police Commissioner [2018] UKSC 40; [2018] 1 WLR 4021.
94
eg Van Colle [106] per Lord Carswell, [139] per Lord Brown.
 Precluding liability for omission in relation to claims brought against public
authorities helps ensure that it is not the courts that are determining the
benefits to which members of the public are entitled.95

Were English law to return to the ‘Anns doctrine’ whereby authorities were
liable for harm foreseeeably resulting from their inaction, liability would become
virtually limitless. Given public authorities’ resources (and ‘immortality’—they
neither die nor become insolvent), they are tempting targets in the litigator’s search
for a defendant able to satisfy judgment. The result would be to place a
disproportionate burden of liability on causally (ie morally?) peripheral parties. 96
When the primary wrongdoer is (as so often) impecunious, the attraction of suing a
public authority which failed to intervene is stronger still. But since the authority
would here have no realistic chance of recovering contribution from the main
tortfeasor, the ‘over-deterrence of peripheral parties’ would be aggravated. Baroness
Hale accepts that wrongdoers’ (anti-social tenants’) inability to pay damages ‘is not
by itself a good enough reason to transfer the liability to someone else’ (local
authority landlords).97
The nonfeasance rule seems a sensible limit on public authority liability. But
not solely on grounds of ‘principle’. A number of policy reasons bolster its retention
—arguably more convincingly than Lord Hoffmann’s ‘autonomy’ rationale in Stovin.
It is important to consider all the reasons which justify the nonfeasance rule
given that it is not absolute. When deciding the width of exceptions to it, it is
necessary to decide the strength and importance of the rule itself. Why ignore the
policy reasons that also support its existence? It is orthodox that positive duties to act
can exceptionally arise by antecedent creation of danger or by the assumption of
responsibility. The court will need to weigh the cogency of the claimant’s argument
for exceptional liability against the rationale for preserving the omissions rule. The
first of the exceptions (creation of danger) is not simply a question of fact but requires
judgement by the court. It is clear enough in a case like Attorney-General of the
British Virgin Islands v Hartwell where the danger created by the defendant was
strong and obvious (indeed proverbial: a loaded gun). 98 According to Lord Hope,
discussing Hartwell, the court must weigh up the level of danger created and the
likelihood of its happening, and the ease with which it could have been averted (in
Hartwell the steps needed, ie the training and supervision of armed police officers,
had not been onerous).99
Whether the defendant did initially create the risk is not always clear. In
Mitchell v Glasgow City Council for example, Lord Hope held that the defender
authority had not created the danger,100 but Lord Rodger thought that it was not a clear
omissions case—not like the hypothetical bystander ‘watching and doing nothing as a
child drowns in a shallow pool or a blind man walks into the path of an oncoming
car’. For Lord Rodger the defenders in Mitchell had been involved in the events
which ‘gave the occasion’ for the pursuer’s fatal assault by a fellow council tenant. 101
95
C Booth & D Squires, The Negligence Liability of Public Authorities (OUP 2006), 3.108.
96
J Stapleton, ‘Duty of care: peripheral parties and alternative opportunities for deterrence’
(1995) 111 LQR 301.
97
Mitchell at [77].
98
[2004] UKPC 12, [2004] 1 WLR 1273.
99
Mitchell at [18].
100
Mitchell at [20].
101
Mitchell at [55].
Baroness Hale expressly agreed with Lord Rodger concluding: ‘it is not quite enough
to say that the complaint of a failure to warn is a complaint of a pure omission. But
the question remains whether it is fair, just and reasonable to impose such a
safeguarding duty upon the local authority’.102
This discussion shows that the existence and scope of the ‘danger creation’
exception may be controversial (and involves weighing up competing factors, not a
bright-line rule). Senior judges can and do differ over whether a case truly concerns a
‘pure omission’ or not, as in Mitchell v Glasgow CC.103 In Robinson itself, the
Supreme Court unanimously held that the police officers were positively ‘acting’ (ie
attempting to arrest the drug dealer) when they collided with the claimant. By
contrast the Court of Appeal had held (rather questionably) that the police had failed
to prevent the drug dealer from injuring the claimant.104
But these difficulties pale compared to the most far-reaching exception to the
nonfeasance rule, the ‘assumption of responsibility’. This concept has received fierce
and powerful academic criticism.105 It might seem sensible enough for tort law to
recognise exceptional liabilities when a defendant expressly assumes them. But
‘assumption of responsibility’ extends way beyond that narrow literal meaning.
Application of the doctrine shows frequent judicial willingness to ‘infer’ such
‘voluntary assumption’. The real question becomes when and why such an inference
will take place: any idea that the defendant has actually, voluntarily accepted
liabilities fades into the background. This is not a harmless fiction. It conceals the
real reasons for the decision, and obfuscates the basis of liability. Like other tort
duties the ones in this area are typically imposed by law—yet they masquerade as
‘voluntarily assumed’.
Notwithstanding such powerful criticisms, assumption of responsibility has
been presented as a magic formula for creating liability which otherwise would not
arise in a number of ‘problem’ areas of tort law. Most prominently as the
rationalisation of the ‘Hedley Byrne principle’ for recovery of pure economic loss;106
perhaps for psychiatric injury;107 more recently as the explanation for ‘non-delegable
duties of care’;108 in addition to conjuring up positive duties to protect others.
Undoubtedly there are some cases where the defendant expressly assumed
responsibility.109 But these are exceptional. In the economic loss cases there has long
been judicial scepticism about the concept’s utility. These run back at least to Lord
Griffiths’ speech in Smith v Bush.110 As recently observed, ‘In effect Lord Griffiths
102
Mitchell at [76].
103
In Robinson at [117] Lord Hughes suggests that the leading omissions case Michael v
Chief Constable could have been analysed as a series of positive acts.
104
[2014] EWCA Civ 15 [53].
105
eg A Robertson and J Wang, ‘The Assumption of Responsibility’ in K Barker, R
Grantham and W Swain (eds), The Law of Misstatements: 50 Years on from Hedley Byrne v
Heller (Hart Publishing, Oxford, 2015).
106
Henderson v Merrett Syndicates Ltd [1995] 2 AC 145.
107
eg Swinney v Chief Constable of Northumbria [1997] QB 464, W v Essex CC [2001] 2 AC
592.
108
Woodland v Swimming Teachers Association [2013] UKSC 66, [2014] AC 537.
109
eg Welsh v Chief Constable of Merseyside [1993] 1 All ER 692 (promise by CPS to inform
court that offences had been ‘taken into consideration’ in other proceedings); Calvert v
William Hill Credit Ltd [2008] EWHC 454 (Ch) (promise to stop problem gambler placing
bets).
110
[1990] 1 AC 831, 862-864.
was suggesting that the [assumption of responsibility] test identified only a conclusion
rather than a criterion’.111 In the leading case of Customs and Excise Commissioners v
Barclays Bank the House of Lords accepted that policy factors were of considerable
importance in determining whether a defendant should be taken ‘objectively’ to have
assumed responsibility. The question is not so much whether the defendant has
actually personally (or ‘subjectively’) assumed responsibility but whether the law
deems him to have done so.
That point has also been recognised in the ‘positive duties to act’ case-law. In
Mitchell v Glasgow CC Lord Scott discussed a South African case in which a
company’s liability for failing to rescue the crew of its stricken fishing boat was
founded ostensibly upon assumption of responsibility. 112 Yet as Lord Scott
observed:113

The company plainly did not regard themselves as having assumed

responsibility for taking steps to rescue the crew from the danger they were in
as a result of a breakdown of the boat’s engine, but Schreiner JA treated them
as having assumed that responsibility. He did so because the boat was the
company’s boat, supplied for use by the crew members for the purposes of the
joint enterprise…

(Perhaps another explanation of this case is that employers stand in a ‘special

relationship’ to employees requiring them to take positive steps to protect them from
harm; but this would equally beg the question.)
In Robinson Lord Hughes again saw the problem. It was not sufficient, he
said, to classify a case as ‘nonfeasance’ to establish that no duty is owed. After all,

The law readily finds [an assumption of responsibility to act] in many

common situations, such as employment, teaching, healthcare and the care of
children, and imposes liability for omitting to protect others. It could equally
readily do so in the case of police officers with a general public duty to protect
the peace, but it does not.114

This brief comment exposes the incoherence of ‘assumption of responsibility’. It is

very readily inferred or implied (or as Lord Hughes bluntly says ‘imposed’) in
situations where the court concludes there should be liability. No particular words or
conduct ‘assuming responsibility’ were identified on the part of medical staff in such
leading cases as Barnett v Chelsea Hospital (positive duty to examine patient
presenting at hospital)115 or Woodland v Essex CC (non-delegable duty owed by
education authority to schoolchildren).116 The truth is that the exceptional duties in
those cases (positive duty to treat; duty to ensure care taken by independent
contractor) arose as a matter of law rather than by any actual ‘assumption of
responsibility’. To pretend that the defendant’s agreement was the basis for liability
evades giving a proper explanation.

111
NRAM Ltd v Steel [2018] UKSC 13, [2018] 1 WLR 1190, [20] per Lord Wilson.
112
Silva Fishing Corporation (Pty) Ltd v Maweza [1957 (2)] SA 256. See Mitchell [43]-[44].
113
Mitchell [54] (emphasis added)
114
Robinson [115]
115
[1969] 1 QB 428.
116
[2014] AC 537.
 The ease with which ‘assumption of responsibility’ is ‘found’ in such cases
may be contrasted (as Lord Hughes notes) with the courts’ consistent refusal to infer
this in the police cases. In Michael for example, the claimant had dialled 999 and the
operator promised to pass her call for help to the police (their response being
significantly delayed by administrative error). More evidence of ‘assumption of
responsibility’ exists here, arguably, than in Barnett or Woodland in which it was
‘inferred’ as a matter of routine. There was at least direct communication between the
parties in Michael. But the majority tersely dismissed the argument that the 999
operator had assumed responsibility on behalf of the police as ‘not tenable’ when she
had not given any assurances about whether and when they would come to the
claimant’s aid.117 In dissent Lord Kerr questioned whether it was right (either in
principle (‘logic’) or in law) to insist that an explicit promise of help must be
present.118 More fundamentally Lord Kerr posed the question (like Lord Hughes in
Robinson) why, when other professionals routinely owe positive duties to their clients
(eg doctors to patients), should the police not? ‘Any other professional would be
liable for inaction with such grievous consequences. So also should be the police’, in
Lord Kerr’s view.119
Notoriously, even the different emergency services seem to be treated
differently. No assumption of responsibility arises when police or fire brigades
actually attend an incident (surely a fortiori from promising to attend),120 yet a mere
promise that an ambulance is on its way generates liability.121 If the rationale of the
third case is that a more specific promise was made by the emergency ambulance call-
handler than by the operator in Michael,122 that does not explain the distinction with
cases of fire engines and police cars actually arriving at the scene. The suggestion is
sometimes made that in Kent v Griffiths the claimant had detrimentally relied upon
the promise (foregoing other opportunities to get to hospital). This seems powerful,
but it is an ex post facto rationalisation not reflected in the reasoning of the Court of
Appeal. Instead Lord Woolf MR drew a close analogy between the ambulance
service and hospitals (where, as seen, positive duties are uncontroversially accepted)
as a way of distinguishing the fire and police cases. 123 This might suggest liability
based on ‘special relationship’ between patient and ambulance service rather than any
actual ‘assumption of responsibility’.124
In conclusion, the nonfeasance rule is far from clear-cut. Its basis has been
criticised. Wide-ranging exceptions exist. The ‘assumption of responsibility’, in
particular, involves policy reasoning—necessarily so since the question is whether to
‘attribute’ an assumption of responsibility to the defendant.125 In the most extensive
judicial discussion of alleged assumption of responsibility by the police (to a covert
human intelligence source), the Court of Appeal divided three ways on how to analyse
117
Michael [138] per Lord Toulson.
118
Michael [164]-[165]. Cf Swinney (above) (‘arguable’ that police had ‘assumed
responsibility’ to keep information confidential, despite submission that this could not be
inferred merely from receipt of information).
119
Michael [181].
120
eg Alexandrou v Oxford [1993] 4 All ER 328 (police), Capital & Counties (fire).
121
Kent v Griffiths [2001] QB 36.
122
Michael [138] per Lord Toulson
123
Kent v Griffiths at [45].
124
The latter phrase is absent from Lord Woolf’s judgment.
125
Caparo Industries plc v Dickman [1990] 2 AC 605, 637 per Lord Oliver. And see
Customs v Barclays (passim).
the liability.126 Arden LJ said that an assumption of responsibility could arise either
from express words, or from ‘taking control of a situation’ (eg detaining someone in a
cell), or otherwise ‘on an evaluation of the facts’.127 Clearly this goes beyond express
assumptions of responsibility.128 Arden LJ held that the police had assumed
responsibility to the informer on the facts. Her Ladyship inferred this in part from
policy reasons favouring the protection of police informers (‘the encouragement of
the free flow of information [about criminal activities] without inhibition’), 129
reinforced by the legislative scheme for the protection of ‘human intelligence
sources’.130 There was full consideration of the ‘Hill principle’ (ie the need to avoid
defensive policing) by all three members of the Court of Appeal. In short, this case
shows that when considering whether the police (or another public authority) have
‘assumed responsibility’ to protect the claimant this requires consideration of all the
factors for and against liability. This undermines the superficially simple distinction
between acts and omissions proffered in Robinson.

Conclusion

Robinson v Chief Constable is a significant case. It usefully reminds tort lawyers of a

number of basic principles. It is only in novel cases that full consideration of all the
reasons for and against duty of care need to be weighed up. Who could doubt it? Yet
most cases which reach the rarefied level of appellate adjudication (albeit these are a
miniscule fraction of tort litigation overall) will by definition raise issues
characteristic of a ‘hard case’. For instance which of two conflicting precedents
should be followed; whether a principle should be refined, extended, or an exception
made. In other words, cases where the duty of care are litigated as a live issue always
raise novel questions to some degree. Precedent is important, but cannot be applied
mechanically. Lord Reed has not gone ‘the full Dworkin’—his Lordship does not
suggest that only ‘principles’ are relevant for (tort) adjudication.131 Lord Reed accepts
that policy may also be relevant, in novel claims. However, such cases are much
more common than he contemplates. So the apparent attempt to limit policy
reasoning may not change much in the end.132
Whether Robinson does herald a radical shift in negligence reasoning depends
on acceptance of Lord Reed’s argument that virtually all we need to explain public
authority liability is the acts-omissions distinction. For Lord Reed that means that the
formerly extensive reliance on public policy is to be sidelined: an unfortunate relict of
the Anns v Merton heresy, now best forgotten. Lord Mance and Lord Hughes were
not persuaded by this revisionism. Understandably so. The nonfeasance principle
needs to be explained and justified, not just postulated as an axiom of liability. It is
dubiously axiomatic when other legal systems have taken different stances towards
(eg) the duty to rescue, and when public authorities have extensive positive duties as a

126
An Informer v A Chief Constable [2012] EWCA Civ 197, [2013] QB 579.
127
Ibid [97]-[100] citing (inter alia) Welsh and Swinney above.
128
Cf Robinson [2014] EWCA Civ 15 [66] per Arnold J: ‘the best explanation of Rigby may
be that it is an assumption of responsibility case’ (see Rigby v Chief Constable above—NB no
express undertakings by the officer ordering use of CS gas canister).
129
An Informer [114], citing Swinney p 487 per Ward LJ.
130
An Informer [115] et seq; Regulation of Investigatory Powers Act 2000, s 29.
131
cf R Dworkin, Taking Rights Seriously (Duckworth 1977).
132
However NB the Robinson ‘established duty’ approach applied in Darnley n 000 above.
matter of public law—not least under article 2 of the ECHR which may lead to
Human Rights Act damages claims.133 Finally, although the simplistic ‘test’ view of
Caparo was rightly disapproved again, it is regrettable that the Robinson court did not
consider criticisms of the vacuity (even the mendacity) of ‘assumption of
responsibility’. Clearly it will continue to feature prominently in nonfeasance claims
against public authorities. It will be another conduit for policy reasoning.
When faced with an arguable point on duty the courts should openly
acknowledge and rigorously analyse all the arguments for and against liability.
History suggests that attempting to subsume these beneath doctrinal concepts does not
ultimately work, but leads to unnecessary formalist gymnastics if judges are to reach
acceptable decisions without trespassing on forbidden ground. Hopefully Robinson
does not point the way into such a dead-end.134

133
eg Michael [139].
134
For post-Robinson policy reasoning in a case accepted to raise a novel extension of
liability cf James-Bowen v Metropolitan Police Commissioner [2018] UKSC 40.